Sensitive Data Mediation for Privacy-Safe ML Requests
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Machine learning models may inadvertently access and store sensitive data, violating confidentiality agreements and data privacy laws, which discourages their use for tasks involving sensitive information.
Innovation Solution
A sensitive data protection service transforms requests containing sensitive data into generic forms before sending them to the machine learning model, and then reverts the response to the original form without the model accessing the sensitive data, using techniques like data removal, replacement, and obfuscation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If machine learning models are provided with access to sensitive data to perform tasks, then task performance efficiency and accuracy are improved, but data privacy and confidentiality are compromised
Solution Approach 1:
The patent introduces a sensitive data protection service as an intermediary layer between the user and the machine learning model. This service receives the original request containing sensitive data, transforms it into a modified request with generic data, forwards it to the model, and then transforms the response back. This intermediary mechanism allows the model to perform tasks efficiently while never directly accessing sensitive data, thus resolving the contradiction between productivity and data privacy.
Solution Approach 2:
The patent creates a copy of the original request with sensitive data replaced by generic data. This copied request is then processed by the machine learning model. The model operates on this copy rather than the original sensitive data, maintaining task performance while preventing privacy violations. The response is then mapped back to the original context without exposing the model to actual sensitive information.
2Object-affected harmful factors
If sensitive data is removed or replaced with generic data before model processing, then data privacy is maintained, but information completeness and task accuracy may deteriorate
Solution Approach 1:
The patent changes the parameter state of the data from sensitive to generic during model processing. By transforming sensitive data into generic data with similar structural properties, the system maintains the informational content needed for task performance while changing the sensitivity parameter. The response is then transformed back, preserving information completeness without exposing sensitive data to the model.
3Object-affected harmful factors
If a data transformation and reversal process is implemented, then data privacy is maintained, but system complexity increases
Solution Approach 1:
The sensitive data protection service performs multiple functions within a single system component: receiving requests, transforming data to generic form, forwarding to model, receiving response, and transforming response back. This multi-functional approach consolidates the complexity into a dedicated service rather than distributing it throughout the entire system, making the complexity manageable and localized.
Data Source
AI summary
Methods and systems for managing sensitive data are disclosed. Data indicative of a request may be received. The data may comprise sensitive information, such as information that a user does not want a machine learning model to access. The data may be transformed into a modified request based on replacing at least one portion of the sensitive information with generic information. A response to the request may be generated based on sending the modified request to the machine learning model. The machine learning model may be configured to generate data indicative of the response to the request without accessing the sensitive information.